Hyperparameter Optimization of CNN for Coffee Berry Disease Classification Using the Artificial Bee Colony Algorithm

Abstract

Indonesia is one of the world's largest coffee producers, with a significant contribution to the global market. However, extreme weather challenges, such as the El Nino phenomenon, have led to a decline in coffee production of up to 30%, affecting the quality and quantity of coffee beans. A major challenge in coffee cultivation is coffee berry diseases, such as the coffee berry borer and coffee berry damage, which can cause up to 60% crop loss. Early detection of these diseases is essential to reduce losses and preserve coffee quality. This study seeks to enhance the performance of a Convolutional Neural Network (CNN) model for coffee berry disease classification by optimizing hyperparameters using the Artificial Bee Colony (ABC) algorithm. The research dataset consists of 2100 images with three categories: Healthy Berry, Berry Borer, and Berry Damage. The research stages include data preprocessing, CNN model design, hyperparameter optimization, training, and model evaluation. The results showed that the application of the ABC algorithm succeeded in significantly improving the accuracy of the CNN model compared to the method without optimization. The accuracy result obtained is 97.14% with an architecture consisting of 3 convolutional layers and 3 fully connected layers. This finding makes a real contribution to the development of meta-heuristic-based optimization techniques for coffee fruit disease classification, as well as supporting efforts to improve coffee quality amid the challenges of global climate change.